docs.page 2.0
Agent-ready Documentation for Developers
6 min read
Agents want to read your docs, too.
In 2021, we launched docs.page so open-source maintainers like us could ship documentation like code. Push Markdown to GitHub and your site is live instantly, with no build steps or hosting debt.

Docs sites were built for human readers. Fast forward to today: LLMs are first-class consumers of documentation too. And without machine ingestion, your project becomes invisible to modern dev tools and their users.
We didn't just need a website. We needed a platform built for the agentic age. That's why we shipped docs.page 2.0.

The open-source dilemma
Today, open-source maintainers face a frustrating compromise when trying to make their docs agent-ready:
- Self-Hosting: You control the documentation tool you use and have free reign over its configuration and feature set. The catch? You are entirely on the hook for building, configuring, and maintaining the code and infrastructure yourself, from custom components, to generating custom
llms.txtfiles to managing search indexing. What starts as free quickly turns into engineering effort and cost. - Premium SaaS platforms: You get a beautiful UI and out-of-the-box LLM optimization without touching a line of config. But these platforms are rigid, metered, and expensive. Basic branding, custom domains, or higher AI usage limits are locked behind steep pricing tiers, and the moment you stop paying their monthly subscription, your entire documentation site goes dark.
We believe that open-source developers shouldn't have to choose between owning their documentation and supporting the modern developer workflow.
Enter docs.page 2.0: docs for humans + agents
We went back to the drawing board and built the platform we wanted for our own open-source projects. docs.page 2.0 still serves documentation asmarkdown from any GitHub branch, however it is now modern and agent-ready - bringing premium UI polish and native AI integrations to open source with no platform markup.
An agent harness for the AI workflow
Instead of bolting a chatbot onto a legacy site, we engineered machine-readability into the core:
- Native MCP (Model Context Protocol): every docs.page site automatically exposes an HTTP MCP endpoint. Developers can add your documentation as a set of tools in Cursor, VS Code, or Claude Code by pointing their client at your MCP URL.
llms.txtout of the box: automaticllms.txtandllms-full.txtgeneration lets LLMs ingest your product context without hallucinating.- Ask AI (bring your own key): an in-site AI chat (⌘I) supporting OpenAI, Anthropic, Google, and xAI. Update your
docs.jsonwith a token returned from our CLI (docs agent create) using your own API key, and manage your own rate limits.

A note on "bring your own key"
Our take: AI tokens are no longer a separate, scary line item - they're a standard part of everyday infrastructure, like compute, databases, or hosting. You already have the accounts. By plugging in the keys you already use, you skip the markup and platform fees traditional SaaS tools add on top, and pay only for what you consume.
Modern interface and intelligent searching
Documentation is the primary interface to your code. 2.0 gives you full control over that experience without the engineering overhead - the same polish used in production by projects like React Native Firebase, Melos, and Widgetbook:
- shadcn/ui presets: modern documentation chrome powered by shadcn/ui. Your branding looks professional out of the box - no enterprise tier required.
- Built-in search: no third-party search widget to configure. Every repository gets built-in full-text search behind a global command palette (⌘K).
- Markdown components: native support for GitHub-flavored alert blockquotes (
> [!NOTE]), Mermaid diagrams, and built-in MDX layout components (Tabs, Steps, Code Groups) with no imports required.

Everything needed to publish great documentation
Your documentation should be content you own, not a site you rent.
- Git publishing: your docs live as Markdown in your own GitHub repository - fully portable, no lock-in. Deploy updates by pushing to your repo, preview by pushing to a branch or PR; no build pipeline.
- Reliable CLI preview:
@docs.page/cli previewuses the exact same MDX bundler as our production edge servers, so what you see on localhost is what deploys. - Drop-in upgrade: already on docs.page? 2.0 serves your existing repo as-is - your current
docs.yamlordocs.jsonkeeps working. New agent features are additive; there's nothing to migrate.
This release also hardens the day-to-day experience for humans, which wouldn't have been possible without the open-source community. To everyone who opened an issue or submitted a PR, thank you. We used your feedback to squash legacy bugs, smooth out edge cases, and land heavily requested features (like native Mermaid support for diagrams, custom global announcement banners, and a completely overhauled local preview mode). You can check out the 2.0 release notes to see the complete list of updates.
Bring your docs into the agentic age
We built docs.page 2.0 because we needed a low-maintenance, high-performance way to handle documentation for the next decade of software. It's easy for humans and agents to author., and easy for humans and agents to consume.
Whether you're starting a new project or upgrading an existing repo, there's no dashboard or infrastructure setup. Push your Markdown to GitHub, and your agent-native documentation is live.
Start publishing today at docs.page → npx @docs.page/cli init
